On Sun, 6 Dec 2020, Tamás Kovács wrote:
Any idea how to remove these files? "OVMS# vfs ls /store/usr
64 12-Sep-1965 06:34 ????????.??? 64 12-Sep-1965 06:34 ????????.??? 64 12-Sep-1965 06:34 ????????.???
The filesystem is corrupted. This happened to me in 2018. Below is an edited thread of messages about this from ovmsdev list. -- Steve From: Stephen Casner <casner@acm.org> I know that I can use "config backup" to back up the config to SD. Does the re-format step mean "module factory reset"? -- Steve From: Michael Balzer <dexter@expeedo.de> Yes. You might consider saving the flash partition for further bug analysis before erasing it: https://docs.google.com/document/d/1q5M9Lb5jzQhJzPMnkMKwy4Es5YK12ACQejX_NWEi... Regards, Michael From: Stephen Casner <casner@acm.org> I will save the flash partition as you suggest. Also, in your message telling about the addition of the config backup/restore function you said the backup covers the directories "ovms_config" and "events" from "/store". I note that /store also contains a .htpasswd file. Should that also be included, or is it already? -- Steve From: Michael Balzer <dexter@expeedo.de> The .htpasswd file is generated automatically by the webserver from the password config. Regards, Michael From: Stephen Casner <casner@acm.org> What kind of <path> is expected in "config backup <path>"? I tried "/sd/config-20181221.zip" and "config-20181221.zip" but both resulted in "Error: zip failed: No such file or directory". Manually creating the empty file "/sd/config-20181221.zip" first did not help. -- Steve From: Michael Balzer <dexter@expeedo.de> I missed the case of a non-existent "/store/events" directory. I'll fix that now, to backup without updating, do "vfs mkdir /store/events". Regards, Michael From: Stephen Casner <casner@acm.org> Michael, Glad you figured that out. I had looked at the code to see if that gave me any hints as to what kind of <path> argument was required, but I could not tell from the code where the error was occurring. I notice that some other commands use <vfspath> for the help string to be more specific. Would that be a good idea for config backup? -- Steve From: Stephen Casner <casner@acm.org> On Fri, 21 Dec 2018, Michael Balzer wrote:
You might consider saving the flash partition for further bug analysis before erasing it https://docs.google.com/document/d/1q5M9Lb5jzQhJzPMnkMKwy4Es5YK12ACQejX_NWEi...
I finally got around to fixing the corrupted /store filesystem on my OVMS v3.0 module. I did save the partition first, as directed by the section of the OVMS v3 Developer Guide to which the reference above points. I also updated the instructions there to add an example of how to mount the saved filesystem on macOS rather than Linux. Looking at the whole filesystem as a raw binary file, I see that the sector at 0x00002000 is all 0xFFs. I don't know if that is what caused the "vfs ls /store" command to show 128 invalid directory entries, but after erasing the flash and rebuilding the /store filesystem and then saving the flash partition to another file on my Mac, there is not a sector of 0xFF like that. Running fsck_msdos on the broken filesystem (as attached on macOS) only shows orphaned clusters, not any invalid directory entries: auge16> fsck_msdos /dev/disk11 ** /dev/rdisk11 ** Phase 1 - Preparing FAT ** Phase 2 - Checking Directories ** Phase 3 - Checking for Orphan Clusters Found orphan cluster(s) Fix? [yn] n Found 25 orphaned clusters 19 files, 800 KiB free (200 clusters) ** Phase 1 - Preparing FAT ** Phase 2 - Checking Directories ** Phase 3 - Checking for Orphan Clusters Found orphan cluster(s) Fix? [yn] n Found 25 orphaned clusters 19 files, 1600 KiB free (400 clusters) ** Phase 1 - Preparing FAT ** Phase 2 - Checking Directories ** Phase 3 - Checking for Orphan Clusters Found orphan cluster(s) Fix? [yn] n Found 25 orphaned clusters 19 files, 2400 KiB free (600 clusters) auge17> -- Steve